TY  - RPRT
TI  - AudioScorer: Learning Question-Aware Token Relevance for Large Audio Language Models
AU  - Marwan Farag
AU  - Bin Yang
PY  - 2026
UR  - https://arxiv.org/abs/2610.26258
ID  - 2610.26258
ER  -